Last Update: Wed Jul 02 11:49:18 +0200 2008

test/spec, a BDD interface for Test::Unit

Copyright (C) 2006, 2007, 2008 Christian Neukirchen <>

What is test/spec?

test/spec layers an RSpec-inspired interface on top of Test::Unit, so you can mix TDD and BDD (Behavior-Driven Development).

test/spec is a clean-room implementation that maps most kinds of Test::Unit assertions to a `should’-like syntax.

Consider this Test::Unit test case:

    class TestFoo < Test::Unit::TestCase
      def test_should_bar
        assert_equal 5, 2 + 3

In test/spec, it looks like this:

    require 'test/spec'

    context "Foo" do
      specify "should bar" do
        (2 + 3).should.equal 5

Since test/spec 0.4, you can also use the new RSpec 1.0 style:

    require 'test/spec'

    describe "Foo" do
      it "should bar" do
        (2 + 3).should.equal 5

test/spec does not include a mocking/stubbing-framework; use whichever you like to. test/spec has been tested successfully with FlexMock and Mocha.

test/spec has no dependencies outside Ruby 1.8.

Mixing test/spec and test/unit

test/spec and Test::Unit contexts/test cases can be intermixed freely, run in the same test and live in the same files. You can just add them to your Rake::TestTask, too. test/spec allows you to leverage your full existing Test::Unit infrastructure.

test/spec does not change Test::Unit with the exception of monkey-patching Test::Unit::TestSuite to order the test cases before running them. (This should not do any harm, but if you know a way around it, please tell me.)

test/spec adds seven global methods, Object#should, Kernel.context, Kernel.xcontext, Kernel.shared_context, Kernel.describe, Kernel.xdescribe, and Kernel.describe_shared. The Kernel methods are private.

You can use assert_* freely in specify-blocks; Object#should works in plain Test::Unit test cases, too, but they will not be counted.

Wrapped assertions

assert_equal:should.equal, should ==
assert_not_equal:should.not.equal, should.not ==
assert_match:should.match, should =~
assert_no_match:should.not.match, should.not =~

(a, an and be without arguments are optional and no-ops.)

Additional assertions

These assertions are not included in Test::Unit, but have been added to test/spec for convenience:

  • should.not.satisfy
  • should.include
  • a.should.predicate (works like assert a.predicate?)
  • operator (where operator is one of >, >=, <, <= or ===)
  • should.output (require test/spec/should-output)

If you write an useful general-purpose assertion, I‘d like to hear of it and may add it to the test/spec distribution.


With more complex assertions, it may be helpful to provide a message to show if the assertion has failed. This can be done with the Should#blaming or Should#messaging methods:

    RUBY_VERSION.should.messaging("Ruby too old.").be > "1.8.4"

    (1 + 1).should.blaming("weird math").not.equal 11

Custom shoulds ("Matchers")

To capture recurring patterns in parts of your specifications, you can define custom "shoulds" (RSpec calls them "matchers") in your contexts, or include modules of them:

    context "Numbers"
      class EqualString < Test::Spec::CustomShould
        def matches?(other)
          object == other.to_s

      def equal_string(str)

      specify "should have to_s"
        42.should equal_string("42")

Alternatively, your implementation can define CustomShould#assumptions, where you can use test/spec assertions instead of Boolean predicates:

    class EqualString < Test::Spec::CustomShould
      def assumptions(other)
        object.should.equal other.to_s

A CustomShould by default takes one argument, which is placed in self.object for your convenience.

You can CustomShould#failure_message to provide a better error message.

SpecDox and RDox

test/spec adds two additional test runners to Test::Unit, based on the console runner but with a different output format.

SpecDox, run with —runner=specdox (or -rs) looks like RSpec‘s output:

    - works for print
    - works for puts
    - works with readline

RDox, run with —runner=rdox (or -rr) can be included for RDoc documentation (e.g. see SPECS):

    == should.output
    * works for print
    * works for puts
    * works with readline

SpecDox and RDox work for Test::Unit too:

    $ ruby -r test/spec test/testunit/test_testresult.rb -rs

    - fault notification
    - passed?
    - result changed notification

    Finished in 0.106647 seconds.

    3 specifications (30 requirements), 0 failures

Since version 0.4, SpecDox and RDox also notice and count empty specifications.

Disabled specifications

Akin to the usual Test::Unit practice, tests quickly can be disabled by replacing specify with xspecify (or it with xit). test/spec will count the disabled tests when you run it with SpecDox or RDox.

When you use xspecify/xit, you also can drop the block. This is useful for writing specifications that you haven‘t yet started implementing.

Complete contexts can be disabled by using xcontext/xdescribe.


Setup/Teardown methods are run in this order:

  * before(:all) in order of definition
  * before(:each)/setup in order of definition
  * specify
  * after(:each)/setup in order of definition
  * before(:each)/setup in order of definition
  * specify
  * after(:each)/setup in order of definition
  * ...
  * after(:all) in order of definition

Please note that before(:all) and after(:all) are run in their own instance, so all instance variables they set are lost(!) and not visible to other specifications. They are e.g. useful for setting up database connections or starting servers.

Shared contexts

Since version 0.9, you can define shared contexts in test/spec using shared_context/describe_shared. These contexts are not executed on their own, but can be included with it_should_behave_like/behaves_like in other contexts. You can use shared contexts to structure suites with many recurring specifications.


Since version 0.2, test/spec features a standalone test runner called specrb. specrb is like an extended version of testrb, Test::Unit‘s test runner, but has additional options. It can be used for plain Test::Unit suites, too.

    $ specrb -a -s -n should.output

    - works for print
    - works for puts
    - works with readline

    Finished in 0.162571 seconds.

    3 specifications (6 requirements), 0 failures

Run specrb —help for the usage.

test/spec on Rails

If you want to specify your Rails applications, you can use the third-party plugin "test/spec on Rails", which can be found at:

It features testing of model validation, redirection, output, HTTP status, template rendering and URL generation.

Installing with RubyGems

Since version 0.3, a Gem of test/spec is available. You can install with:

    gem install test-spec

I also provide a local mirror of the gems (and development snapshots) at my site:

    gem install test-spec --source


  • September 29th, 2006: First public release 0.1.
  • October 18th, 2006: Second public release 0.2.
    • Better, module-based implementation
    • Official support for FlexMock and Mocha
    • More robust Should#output
    • Should#operator
    • Nested contexts
    • Standalone test/spec runner, specrb
  • January 24th, 2007: Third public release 0.3.
    • should.be_close, should.be_an_instance_of, should.be_a_kind_of, and should.be_nil have been deprecated. Use the dot-variants of them. These assertions will be removed in 1.0.
    • specrb -a now includes -Ilib by default for easier out-of-the-box testing.
    • Added custom shoulds.
    • Added messaging/blaming.
    • Added disabling of specifications.
    • Small bug fixes.
    • Gem available.
  • June 29th, 2007: Fourth public release 0.4.
    • Support for Ruby 1.8.6.
    • Support describe/it/before/after RSpec 1.0 syntax.
    • Allow should.raise { code_that_raises }
    • Add xcontext to disable complete contexts.
    • Backtraces are cleaner now.
    • Mention test/spec on Rails.
    • Fix small Gem bugs.
    • Fix bug related to counting negated assertions.
    • Fix bug in specrb.
    • Allow empty xspecifys.
    • Make SpecDox and RDox count empty specifications.
    • Allow Kernel#context to take a superclass.
  • July 2nd, 2008: Fifth public release 0.9.
    • Allow should.<predicate>? as well as should.<predicate>.
    • Add shared contexts.
    • Nested contexts now run the setups/teardowns/before(:all)/after(:all) of their parents.


Please mail bugs, suggestions and patches to <>.

Darcs repository ("darcs send" is welcome for patches):

Thanks to

  • Eero Saynatkari for writing should.output.
  • Tuxie for writing test/spec on Rails.
  • Brian Donovan for allowing alternative superclasses.
  • Xavier Shay for implementing nested setups/teardowns.
  • Chris Wanstrath for should.raise with a block and xcontext.
  • Jean-Michel Garnier for packaging the first gem.
  • Mikko Lehtonen, Jan Wikholm, Matt Mower and Michael Fellinger for testing the gem.
  • Chris McGrath for reporting a bug.
  • Thomas Fuchs for BDD testing which convinced me.
  • Dave Astels for BDD.
  • The RSpec team for API inspiration.
  • Nathaniel Talbott for Test::Unit.


Copyright (C) 2006, 2007, 2008 Christian Neukirchen <>

test/spec is licensed under the same terms as Ruby itself.

Please mail bugs, feature requests or patches to the mail addresses found above or use IRC to contact the developer.


Behavior-Driven Development:<>
RSpec:<> testing:<>
Christian Neukirchen:<>